Errata and changelog
Every correction, retirement and bank update, newest first. Question IDs never change, so you can match them to your review notes.
| Date | Exam | Question IDs | Change |
|---|---|---|---|
| 2026-10-18 | AIF-C01 | aif-d1-027 to aif-d1-039, aif-d2-038 to aif-d2-053, aif-d3-040 to aif-d3-057, aif-d4-024 to aif-d4-032, aif-d5-020 to aif-d5-028 | Published expansion batch S1 (65 questions) after live citation re-check (65/65), independent review X3 (65/65), and X4 re-check of the 4 items changed after X3 (4/4). Bank grew from 130 to 195. Domain weights remain within ±2 points of the official guide. |
| 2026-10-18 | AIF-C01 | aif-d4-019 | Before launch: clarified the transparency question so it no longer hints at explainability, and replaced two weak distractors (Robustness and Privacy became Controllability and Veracity). Answer and source unchanged. |
| 2026-10-18 | AIF-C01 | aif-d1-013, aif-d1-014, aif-d1-015, aif-d1-022, aif-d1-023, aif-d4-008, aif-d4-009 | Before launch: re-cited to current official pages (SageMaker AI documentation and the AWS overfitting explainer) instead of legacy Amazon Machine Learning documentation. Questions and answers unchanged. |
| 2026-10-18 | AIF-C01 | aif-d3-003 | Before launch: tightened the Top P answer wording to match the documentation (top P% of the cumulative probability distribution). |
| 2026-10-18 | AIF-C01 | aif-d2-022, aif-d2-023, aif-d2-024, aif-d2-026 (retired); aif-d2-034 to aif-d2-037 (new) | Before launch: retired items that depended on time-sensitive or closed offerings (Amazon Q Developer IDE plugin end-of-support notice, Amazon Q Business no longer open to new customers) or that the independent reviewer flagged for scope. Replaced with questions on batch inference, Bedrock data privacy, generative AI limitations, and business value metrics. |
| 2026-10-18 | AIF-C01 | aif-d2-010 (retired), aif-d2-033 (new) | Before launch: the pricing question cited a pricing-table fragment. Replaced with a question on how input tokens affect request cost, citing the Amazon Bedrock CountTokens documentation. |
| 2026-10-18 | AIF-C01 | aif-d3-015, aif-d3-016 (retired); aif-d3-038, aif-d3-039 (new) | Before launch: rewrote the prompt-caching question with a single, fully supported claim; replaced the Amazon Bedrock Agents (Classic) question, now in maintenance mode, with a question on the agent planning module. |
| 2026-10-18 | AIF-C01 | aif-d4-010, aif-d4-011, aif-d4-012, aif-d4-017, aif-d4-018 (retired); aif-d4-019 to aif-d4-023 (new) | Before launch: retired questions citing SageMaker Clarify and Amazon A2I pages, which are no longer open to new customers. Replaced with questions on transparency, explainable AI benefits, labeling AI-generated content, human review safeguards, and intellectual property risk. |
| 2026-10-18 | AIF-C01 | aif-d5-011 (retired), aif-d5-019 (new) | Before launch: retired a question about Amazon GuardDuty, which is on the exam guide's out-of-scope list. Replaced with a question on AWS Secrets Manager. |
| 2026-10-18 | AIF-C01 | aif-d2-028 (retired), aif-d2-032 (new) | Before launch: retired aif-d2-028 because it said custom Amazon Bedrock models can only be used with Provisioned Throughput; current documentation also describes on-demand deployment of custom models. Replaced with aif-d2-032, which tests the cost tradeoff between on-demand and Provisioned Throughput. |
| 2026-10-18 | AIF-C01 | all | First bank: 130 questions mapped to exam guide version 1.1 (published 2026-04-30). |